BPF encap ops can return different types of positive values, such like NET_RX_DROP, NET_XMIT_CN, NETDEV_TX_BUSY, and so on, from function skb_do_redirect and bpf_lwt_xmit_reroute. At the xmit hook, such return values would be treated implicitly as LWTUNNEL_XMIT_CONTINUE in ip(6)_finish_output2. When this happens, skbs that have been freed would continue to the neighbor subsystem, causing use-after-free bug and kernel crashes.

To fix the incorrect behavior, skb_do_redirect return values can be simply discarded, the same as tc-egress behavior. On the other hand, bpf_lwt_xmit_reroute returns useful errors to local senders, e.g. PMTU information. Thus convert its return values to avoid the conflict with LWTUNNEL_XMIT_CONTINUE.

Fixes: 3a0af8fd61f9 ("bpf: BPF for lightweight tunnel infrastructure") Suggested-by: Martin KaFai Lau martin.lau@linux.dev Suggested-by: Stanislav Fomichev sdf@google.com Reported-by: Jordan Griege jgriege@cloudflare.com Signed-off-by: Yan Zhai yan@cloudflare.com --- * v5: discards skb_do_redirect return instead; convert bpf_lwt_xmit_reroute return; * v4: minor commit message changes * v3: converts skb_do_redirect statuses from both ingress and egress * v2: code style amend --- net/core/lwt_bpf.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/net/core/lwt_bpf.c b/net/core/lwt_bpf.c index 8b6b5e72b217..4a0797f0a154 100644 --- a/net/core/lwt_bpf.c +++ b/net/core/lwt_bpf.c @@ -60,9 +60,8 @@ static int run_lwt_bpf(struct sk_buff *skb, struct bpf_lwt_prog *lwt, ret = BPF_OK; } else { skb_reset_mac_header(skb); - ret = skb_do_redirect(skb); - if (ret == 0) - ret = BPF_REDIRECT; + skb_do_redirect(skb); + ret = BPF_REDIRECT; } break;

@@ -255,7 +254,7 @@ static int bpf_lwt_xmit_reroute(struct sk_buff *skb)

err = dst_output(dev_net(skb_dst(skb)->dev), skb->sk, skb); if (unlikely(err)) - return err; + return net_xmit_errno(err);

/* ip[6]_finish_output2 understand LWTUNNEL_XMIT_DONE */ return LWTUNNEL_XMIT_DONE;

There is no lwt_xmit test case for BPF_REDIRECT yet. Add test cases for both normal and abnormal situations. For abnormal test cases, devices are set down or have its carrier set down. Without proper fixes, BPF_REDIRECT to either ingress or egress of such device would panic the kernel.

Signed-off-by: Yan Zhai yan@cloudflare.com --- v5->v6: busybox does not support ping mark, redirect via IP address instead. Add missing kernel test config. --- tools/testing/selftests/bpf/config | 1 + .../selftests/bpf/prog_tests/lwt_helpers.h | 139 ++++++++ .../selftests/bpf/prog_tests/lwt_redirect.c | 330 ++++++++++++++++++ .../selftests/bpf/progs/test_lwt_redirect.c | 90 +++++ 4 files changed, 560 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/lwt_helpers.h create mode 100644 tools/testing/selftests/bpf/prog_tests/lwt_redirect.c create mode 100644 tools/testing/selftests/bpf/progs/test_lwt_redirect.c

diff --git a/tools/testing/selftests/bpf/config b/tools/testing/selftests/bpf/config index 3b350bc31343..fa083f1e5b34 100644 --- a/tools/testing/selftests/bpf/config +++ b/tools/testing/selftests/bpf/config @@ -16,6 +16,7 @@ CONFIG_CRYPTO_USER_API_HASH=y CONFIG_DEBUG_INFO=y CONFIG_DEBUG_INFO_BTF=y CONFIG_DEBUG_INFO_DWARF4=y +CONFIG_DUMMY=y CONFIG_DYNAMIC_FTRACE=y CONFIG_FPROBE=y CONFIG_FTRACE_SYSCALLS=y diff --git a/tools/testing/selftests/bpf/prog_tests/lwt_helpers.h b/tools/testing/selftests/bpf/prog_tests/lwt_helpers.h new file mode 100644 index 000000000000..61333f2a03f9 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/lwt_helpers.h @@ -0,0 +1,139 @@ +/* SPDX-License-Identifier: GPL-2.0 */ + +#ifndef __LWT_HELPERS_H +#define __LWT_HELPERS_H + +#include <time.h> +#include <net/if.h> +#include <linux/if_tun.h> +#include <linux/icmp.h> + +#include "test_progs.h" + +#define log_err(MSG, ...) \ + fprintf(stderr, "(%s:%d: errno: %s) " MSG "\n", \ + __FILE__, __LINE__, strerror(errno), ##__VA_ARGS__) + +#define RUN_TEST(name) \ + ({ \ + if (test__start_subtest(#name)) \ + if (ASSERT_OK(netns_create(), "netns_create")) { \ + struct nstoken *token = open_netns(NETNS); \ + if (ASSERT_OK_PTR(token, "setns")) { \ + test_ ## name(); \ + close_netns(token); \ + } \ + netns_delete(); \ + } \ + }) + +#define NETNS "ns_lwt" + +static inline int netns_create(void) +{ + return system("ip netns add " NETNS); +} + +static inline int netns_delete(void) +{ + return system("ip netns del " NETNS ">/dev/null 2>&1"); +} + +static int open_tuntap(const char *dev_name, bool need_mac) +{ + int err = 0; + struct ifreq ifr; + int fd = open("/dev/net/tun", O_RDWR); + + if (!ASSERT_GT(fd, 0, "open(/dev/net/tun)")) + return -1; + + ifr.ifr_flags = IFF_NO_PI | (need_mac ? IFF_TAP : IFF_TUN); + memcpy(ifr.ifr_name, dev_name, IFNAMSIZ); + + err = ioctl(fd, TUNSETIFF, &ifr); + if (!ASSERT_OK(err, "ioctl(TUNSETIFF)")) { + close(fd); + return -1; + } + + err = fcntl(fd, F_SETFL, O_NONBLOCK); + if (!ASSERT_OK(err, "fcntl(O_NONBLOCK)")) { + close(fd); + return -1; + } + + return fd; +} + +#define ICMP_PAYLOAD_SIZE 100 + +/* Match an ICMP packet with payload len ICMP_PAYLOAD_SIZE */ +static int __expect_icmp_ipv4(char *buf, ssize_t len) +{ + struct iphdr *ip = (struct iphdr *)buf; + struct icmphdr *icmp = (struct icmphdr *)(ip + 1); + ssize_t min_header_len = sizeof(*ip) + sizeof(*icmp); + + if (len < min_header_len) + return -1; + + if (ip->protocol != IPPROTO_ICMP) + return -1; + + if (icmp->type != ICMP_ECHO) + return -1; + + return len == ICMP_PAYLOAD_SIZE + min_header_len; +} + +typedef int (*filter_t) (char *, ssize_t); + +/* wait_for_packet - wait for a packet that matches the filter + * + * @fd: tun fd/packet socket to read packet + * @filter: filter function, returning 1 if matches + * @timeout: timeout to wait for the packet + * + * Returns 1 if a matching packet is read, 0 if timeout expired, -1 on error. + */ +static int wait_for_packet(int fd, filter_t filter, struct timeval *timeout) +{ + char buf[4096]; + int max_retry = 5; /* in case we read some spurious packets */ + fd_set fds; + + FD_ZERO(&fds); + while (max_retry--) { + /* Linux modifies timeout arg... So make a copy */ + struct timeval copied_timeout = *timeout; + ssize_t ret = -1; + + FD_SET(fd, &fds); + + ret = select(1 + fd, &fds, NULL, NULL, &copied_timeout); + if (ret <= 0) { + if (errno == EINTR) + continue; + else if (errno == EAGAIN || ret == 0) + return 0; + + log_err("select failed"); + return -1; + } + + ret = read(fd, buf, sizeof(buf)); + + if (ret <= 0) { + log_err("read(dev): %ld", ret); + return -1; + } + + if (filter && filter(buf, ret) > 0) + return 1; + } + + return 0; +} + +#endif /* __LWT_HELPERS_H */ diff --git a/tools/testing/selftests/bpf/prog_tests/lwt_redirect.c b/tools/testing/selftests/bpf/prog_tests/lwt_redirect.c new file mode 100644 index 000000000000..59b38569f310 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/lwt_redirect.c @@ -0,0 +1,330 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause + +/* + * Test suite of lwt_xmit BPF programs that redirect packets + * The file tests focus not only if these programs work as expected normally, + * but also if they can handle abnormal situations gracefully. + * + * WARNING + * ------- + * This test suite may crash the kernel, thus should be run in a VM. + * + * Setup: + * --------- + * All tests are performed in a single netns. Two lwt encap routes are setup for + * each subtest: + * + * ip route add 10.0.0.0/24 encap bpf xmit <obj> sec "<ingress_sec>" dev link_err + * ip route add 20.0.0.0/24 encap bpf xmit <obj> sec "<egress_sec>" dev link_err + * + * Here <obj> is statically defined to test_lwt_redirect.bpf.o, and each section + * of this object holds a program entry to test. The BPF object is built from + * progs/test_lwt_redirect.c. We didn't use generated BPF skeleton since the + * attachment for lwt programs are not supported by libbpf yet. + * + * For testing, ping commands are run in the test netns: + * + * ping 10.0.0.<ifindex> -c 1 -w 1 -s 100 + * ping 20.0.0.<ifindex> -c 1 -w 1 -s 100 + * + * Scenarios: + * -------------------------------- + * 1. Redirect to a running tap/tun device + * 2. Redirect to a down tap/tun device + * 3. Redirect to a vlan device with lower layer down + * + * Case 1, ping packets should be received by packet socket on target device + * when redirected to ingress, and by tun/tap fd when redirected to egress. + * + * Case 2,3 are considered successful as long as they do not crash the kernel + * as a regression. + * + * Case 1,2 use tap device to test redirect to device that requires MAC + * header, and tun device to test the case with no MAC header added. + */ +#include <sys/socket.h> +#include <net/if.h> +#include <linux/if_ether.h> +#include <linux/if_packet.h> +#include <linux/if_tun.h> +#include <linux/icmp.h> +#include <arpa/inet.h> +#include <unistd.h> +#include <errno.h> +#include <stdbool.h> +#include <stdlib.h> + +#include "lwt_helpers.h" +#include "test_progs.h" +#include "network_helpers.h" + +#define BPF_OBJECT "test_lwt_redirect.bpf.o" +#define INGRESS_SEC(need_mac) ((need_mac) ? "redir_ingress" : "redir_ingress_nomac") +#define EGRESS_SEC(need_mac) ((need_mac) ? "redir_egress" : "redir_egress_nomac") +#define LOCAL_SRC "10.0.0.1" +#define CIDR_TO_INGRESS "10.0.0.0/24" +#define CIDR_TO_EGRESS "20.0.0.0/24" + +/* ping to redirect toward given dev, with last byte of dest IP being the target + * device index. + * + * Note: ping command inside BPF-CI is busybox version, so it does not have certain + * function, such like -m option to set packet mark. + */ +static void ping_dev(const char *dev, bool is_ingress) +{ + int link_index = if_nametoindex(dev); + char ip[256]; + + if (!ASSERT_GE(link_index, 0, "if_nametoindex")) + return; + + if (is_ingress) + snprintf(ip, sizeof(ip), "10.0.0.%d", link_index); + else + snprintf(ip, sizeof(ip), "20.0.0.%d", link_index); + + /* We won't get a reply. Don't fail here */ + SYS_NOFAIL("ping %s -c1 -W1 -s %d >/dev/null 2>&1", + ip, ICMP_PAYLOAD_SIZE); +} + +static int new_packet_sock(const char *ifname) +{ + int err = 0; + int ignore_outgoing = 1; + int ifindex = -1; + int s = -1; + + s = socket(AF_PACKET, SOCK_RAW, 0); + if (!ASSERT_GE(s, 0, "socket(AF_PACKET)")) + return -1; + + ifindex = if_nametoindex(ifname); + if (!ASSERT_GE(ifindex, 0, "if_nametoindex")) { + close(s); + return -1; + } + + struct sockaddr_ll addr = { + .sll_family = AF_PACKET, + .sll_protocol = htons(ETH_P_IP), + .sll_ifindex = ifindex, + }; + + err = bind(s, (struct sockaddr *)&addr, sizeof(addr)); + if (!ASSERT_OK(err, "bind(AF_PACKET)")) { + close(s); + return -1; + } + + /* Use packet socket to capture only the ingress, so we can distinguish + * the case where a regression that actually redirects the packet to + * the egress. + */ + err = setsockopt(s, SOL_PACKET, PACKET_IGNORE_OUTGOING, + &ignore_outgoing, sizeof(ignore_outgoing)); + if (!ASSERT_OK(err, "setsockopt(PACKET_IGNORE_OUTGOING)")) { + close(s); + return -1; + } + + err = fcntl(s, F_SETFL, O_NONBLOCK); + if (!ASSERT_OK(err, "fcntl(O_NONBLOCK)")) { + close(s); + return -1; + } + + return s; +} + +static int expect_icmp(char *buf, ssize_t len) +{ + struct ethhdr *eth = (struct ethhdr *)buf; + + if (len < (ssize_t)sizeof(*eth)) + return -1; + + if (eth->h_proto == htons(ETH_P_IP)) + return __expect_icmp_ipv4((char *)(eth + 1), len - sizeof(*eth)); + + return -1; +} + +static int expect_icmp_nomac(char *buf, ssize_t len) +{ + return __expect_icmp_ipv4(buf, len); +} + +static void send_and_capture_test_packets(const char *test_name, int tap_fd, + const char *target_dev, bool need_mac) +{ + int psock = -1; + struct timeval timeo = { + .tv_sec = 0, + .tv_usec = 250000, + }; + int ret = -1; + + filter_t filter = need_mac ? expect_icmp : expect_icmp_nomac; + + ping_dev(target_dev, false); + + ret = wait_for_packet(tap_fd, filter, &timeo); + if (!ASSERT_EQ(ret, 1, "wait_for_epacket")) { + log_err("%s egress test fails", test_name); + goto out; + } + + psock = new_packet_sock(target_dev); + ping_dev(target_dev, true); + + ret = wait_for_packet(psock, filter, &timeo); + if (!ASSERT_EQ(ret, 1, "wait_for_ipacket")) { + log_err("%s ingress test fails", test_name); + goto out; + } + +out: + if (psock >= 0) + close(psock); +} + +static int setup_redirect_target(const char *target_dev, bool need_mac) +{ + int target_index = -1; + int tap_fd = -1; + + tap_fd = open_tuntap(target_dev, need_mac); + if (!ASSERT_GE(tap_fd, 0, "open_tuntap")) + goto fail; + + target_index = if_nametoindex(target_dev); + if (!ASSERT_GE(target_index, 0, "if_nametoindex")) + goto fail; + + SYS(fail, "ip link add link_err type dummy"); + SYS(fail, "ip link set lo up"); + SYS(fail, "ip addr add dev lo " LOCAL_SRC "/32"); + SYS(fail, "ip link set link_err up"); + SYS(fail, "ip link set %s up", target_dev); + + SYS(fail, "ip route add %s dev link_err encap bpf xmit obj %s sec %s", + CIDR_TO_INGRESS, BPF_OBJECT, INGRESS_SEC(need_mac)); + + SYS(fail, "ip route add %s dev link_err encap bpf xmit obj %s sec %s", + CIDR_TO_EGRESS, BPF_OBJECT, EGRESS_SEC(need_mac)); + + return tap_fd; + +fail: + if (tap_fd >= 0) + close(tap_fd); + return -1; +} + +static void test_lwt_redirect_normal(void) +{ + const char *target_dev = "tap0"; + int tap_fd = -1; + bool need_mac = true; + + tap_fd = setup_redirect_target(target_dev, need_mac); + if (!ASSERT_GE(tap_fd, 0, "setup_redirect_target")) + return; + + send_and_capture_test_packets(__func__, tap_fd, target_dev, need_mac); + close(tap_fd); +} + +static void test_lwt_redirect_normal_nomac(void) +{ + const char *target_dev = "tun0"; + int tap_fd = -1; + bool need_mac = false; + + tap_fd = setup_redirect_target(target_dev, need_mac); + if (!ASSERT_GE(tap_fd, 0, "setup_redirect_target")) + return; + + send_and_capture_test_packets(__func__, tap_fd, target_dev, need_mac); + close(tap_fd); +} + +/* This test aims to prevent regression of future. As long as the kernel does + * not panic, it is considered as success. + */ +static void __test_lwt_redirect_dev_down(bool need_mac) +{ + const char *target_dev = "tap0"; + int tap_fd = -1; + + tap_fd = setup_redirect_target(target_dev, need_mac); + if (!ASSERT_GE(tap_fd, 0, "setup_redirect_target")) + return; + + SYS(out, "ip link set %s down", target_dev); + ping_dev(target_dev, true); + ping_dev(target_dev, false); + +out: + close(tap_fd); +} + +static void test_lwt_redirect_dev_down(void) +{ + __test_lwt_redirect_dev_down(true); +} + +static void test_lwt_redirect_dev_down_nomac(void) +{ + __test_lwt_redirect_dev_down(false); +} + +/* This test aims to prevent regression of future. As long as the kernel does + * not panic, it is considered as success. + */ +static void test_lwt_redirect_dev_carrier_down(void) +{ + const char *lower_dev = "tap0"; + const char *vlan_dev = "vlan100"; + int tap_fd = -1; + + tap_fd = setup_redirect_target(lower_dev, true); + if (!ASSERT_GE(tap_fd, 0, "setup_redirect_target")) + return; + + SYS(out, "ip link add vlan100 link %s type vlan id 100", lower_dev); + SYS(out, "ip link set %s up", vlan_dev); + SYS(out, "ip link set %s down", lower_dev); + ping_dev(vlan_dev, true); + ping_dev(vlan_dev, false); + +out: + close(tap_fd); +} + +static void *test_lwt_redirect_run(void *arg) +{ + netns_delete(); + RUN_TEST(lwt_redirect_normal); + RUN_TEST(lwt_redirect_normal_nomac); + RUN_TEST(lwt_redirect_dev_down); + RUN_TEST(lwt_redirect_dev_down_nomac); + RUN_TEST(lwt_redirect_dev_carrier_down); + return NULL; +} + +void test_lwt_redirect(void) +{ + pthread_t test_thread; + int err; + + /* Run the tests in their own thread to isolate the namespace changes + * so they do not affect the environment of other tests. + * (specifically needed because of unshare(CLONE_NEWNS) in open_netns()) + */ + err = pthread_create(&test_thread, NULL, &test_lwt_redirect_run, NULL); + if (ASSERT_OK(err, "pthread_create")) + ASSERT_OK(pthread_join(test_thread, NULL), "pthread_join"); +} diff --git a/tools/testing/selftests/bpf/progs/test_lwt_redirect.c b/tools/testing/selftests/bpf/progs/test_lwt_redirect.c new file mode 100644 index 000000000000..7ab1fd310efb --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_lwt_redirect.c @@ -0,0 +1,90 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <linux/bpf.h> +#include <linux/ip.h> +#include <bpf/bpf_endian.h> +#include <bpf/bpf_helpers.h> +#include "bpf_tracing_net.h" + +/* We don't care about whether the packet can be received by network stack. + * Just care if the packet is sent to the correct device at correct direction + * and not panic the kernel. + */ +static int prepend_dummy_mac(struct __sk_buff *skb) +{ + char mac[] = {0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0xf, + 0xe, 0xd, 0xc, 0xb, 0xa, 0x08, 0x00}; + + if (bpf_skb_change_head(skb, ETH_HLEN, 0)) + return -1; + + if (bpf_skb_store_bytes(skb, 0, mac, sizeof(mac), 0)) + return -1; + + return 0; +} + +/* Use the last byte of IP address to redirect the packet */ +static int get_redirect_target(struct __sk_buff *skb) +{ + struct iphdr *iph = NULL; + void *start = (void *)(long)skb->data; + void *end = (void *)(long)skb->data_end; + + if (start + sizeof(*iph) > end) + return -1; + + iph = (struct iphdr *)start; + return bpf_ntohl(iph->daddr) & 0xff; +} + +SEC("redir_ingress") +int test_lwt_redirect_in(struct __sk_buff *skb) +{ + int target = get_redirect_target(skb); + + if (target < 0) + return BPF_OK; + + if (prepend_dummy_mac(skb)) + return BPF_DROP; + + return bpf_redirect(target, BPF_F_INGRESS); +} + +SEC("redir_egress") +int test_lwt_redirect_out(struct __sk_buff *skb) +{ + int target = get_redirect_target(skb); + + if (target < 0) + return BPF_OK; + + if (prepend_dummy_mac(skb)) + return BPF_DROP; + + return bpf_redirect(target, 0); +} + +SEC("redir_egress_nomac") +int test_lwt_redirect_out_nomac(struct __sk_buff *skb) +{ + int target = get_redirect_target(skb); + + if (target < 0) + return BPF_OK; + + return bpf_redirect(target, 0); +} + +SEC("redir_ingress_nomac") +int test_lwt_redirect_in_nomac(struct __sk_buff *skb) +{ + int target = get_redirect_target(skb); + + if (target < 0) + return BPF_OK; + + return bpf_redirect(target, BPF_F_INGRESS); +} + +char _license[] SEC("license") = "GPL";